Atlas

Models

← All models

Gemini 2.5 Pro Preview 03-25

Google DeepMind · 2025-04-04

Google released `gemini-2.5-pro-preview-03-25` on April 4, 2025 as the billing-enabled public-preview endpoint for the 03-25 Gemini 2.5 Pro iteration, while `gemini-2.5-pro-exp-03-25` remained available on the free tier. Google later redirected this endpoint to newer Gemini 2.5 Pro versions, so this row denotes its original 03-25 deployment.

Benchmark scores

BenchmarkScore
Agentic Tool Use (Chat)62.4
Agentic Tool Use (Enterprise)68.8
AidanBench3212
Aider Polyglot72.9
ARC-AGI-133.0
ARC-AGI-23.8
BuzzBench71.1
CadEval64.0
Capability146.1
Confabulations Leaderboard Confabulation Rate4.0
Confabulations Leaderboard Non-Response Rate17.6
Confabulations Leaderboard Weighted Score10.8
Elimination Game TrueSkill Mu4.5
Epoch Capabilities Index (ECI)145
EQ-Bench 31393
EQ-Bench 3 (Rubric Score)67.3
EQ-Bench 3: Analytic8.2
EQ-Bench 3: Assertive4.0
EQ-Bench 3: Compliant7.2
EQ-Bench 3: Empathy7.5
EQ-Bench 3: Humanlike6.1
EQ-Bench 3: Insight7.3
EQ-Bench 3: Moralising2.6
EQ-Bench 3: Pragmatic7.1
EQ-Bench 3: Safety5.8
EQ-Bench 3: Social IQ6.5
EQ-Bench 3: Warm7.6
Fiction.liveBench66.7
ForecastBench Baseline Dataset Brier Index61.4
ForecastBench Baseline Market Brier Index60.8
Loading Atlas data…